aliasing

[arily]

Hi,
I have noticed that the upmixed result contains aliasing noises.

Here's the original stereo audio (44.1khz oversampled to 88.2 to demonstrate)

matrix result: (4.0 quad)

as you can see both above f/2 and below is not as clean as the original signal.

I noticed adjusting fft window size changes the color of the aliases.

I also tried using f64 and lower amplitude to leave headrooms, prevent clipping, but these methods only reduces the noise, not eliminating it.

[GWBasic]

Sorry for the delay.

Looking at the above images, I'm not sure what you mean by aliasing. The output of de-matrixing is never going to be as good as having something mixed to discrete quad; it will always be a best-effort approximation. Some artifacts will always be present; if you want something perfect you'll need to go back to master tapes or use AI stemming.

There will be no clipping with default options when outputting quad. (Clipping can happen when derriving a center channel or subwoofer with the -loud option.)

f64 will not resolve aliasing, only quantization issues. Generally, to resolve aliasing issues, use a higher sampling rate or otherwise you'll need to understand sample processing better than I do.

---

Can you provide very clear steps to reproduce? IE, provide a short input wave file that has test tones and the command-line arguments that generate aliasing?